The Research Of Embedding-based Knowledge Construction And Recommendation

With the rapid development of Internet technology,massive information is released on the Internet,and we came to the era of information explosion.There is a wealth of knowledge behind information.Knowledge graph expresses information in a form which is closer to the human cognitive world.How to efficiently acquire and utilize knowledge from large-scale knowledge graphs is an important research topic,which mainly includes two aspects:knowledge construction and acquisition.At present,the related research works are based on discrete symbolic representation,large-scale and efficient calculations cannot be performed,and implicit semantic cannot be utilized.Therefore,it is a challenge to improve the efficiency and quality of knowledge construction and acquisition.This paper proposes a knowledge construction and acquisition scheme based on knowledge graph embedding technology,which can express symbolized knowledge as vectors and convert semantic relationships into numerical calculations.Firstly,this paper proposes a knowledge construction method and a universal framework named OWLearner,which is used to construct the ontology layer above the data level of the knowledge graph,i.e.,the pattern layer,and use OWL ontology to describe the knowledge.This paper has learned 12 OWL ontology axiom patterns.The training sample and the feature construction method are given.Secondly,this paper proposes a knowledge acquisition mechanism,focusing on the recommendation of knowledge,and proposes a universal recommendation framework TrQuery for SPARQL query.The query parser algorithm,semantic-based scoring model and recommendation algorithm are given.Finally,a number of experiments were designed to evaluate the correctness and performance of OWLearner and TrQuery.Experiments show that OWLearner and TrQuery are superior to the traditional methods in terms of both time efficiency and effectiveness.In summary,this paper proposes an adaptive knowledge construction framework OWLearner based on embedding technology and a knowledge acquisition framework TrQuery,both of which can deal with large-scale knowledge graph effectively and conveniently.